ebook img

(in)consistency and real options: Much ado about nothing? PDF

19 Pages·2016·0.3 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview (in)consistency and real options: Much ado about nothing?

Time (in)consistency and real options: Much ado about nothing? Josef Schosser∗ January 2016 Abstract If decision makers have exclusive rights to particular investment pro- jects, they frequently have the opportunity to delay these investments. This paper analyzes the effect of quasi-hyperbolic discounting, i.e. time- inconsistent preferences on the exercise timing of such options to defer an investment. It complements earlier work on this issue covering risk aversion and capital market interaction. The results are as follows: In a number of cases, the capital market environment provides for the irrele- vance of quasi-hyperbolic discounting. Besides this, a different behavior of time-inconsistent and time-consistent decision makers occurs only for quite specific parameter conditions. In light of experimental evidence for time-inconsistent behavior, this provokes the following question: Is time- inconsistent behavior really driven by quasi-hyperbolic discounting, but rather by more fundamental irrationality (like a disregard of fairly ubiq- uitous market opportunities)? Keywords: Capital market interaction, Quasi-hyperbolic discounting, Real options, Risk attitude, Time inconsistency. JEL Classification: D81, D91, G13, G31. ∗UniversityofPassau,DepartmentofBusinessAdministrationandEconomics,94030Pas- sau,Tel.+498515093274,e-mail: [email protected]. 1 1 Introduction Timeinconsistency(sometimescalleddynamicinconsistency)isamongthemost extensively discussed behavioral departures of decision theoretic directives. In the words of Lengwiler (2004, p. 145), it can be described as follows: “We say that an ... intertemporal decision is time consistent (emphasis in original) if, when the next period comes along, it will be optimal for you to stick to the same planned path. The decision is time inconsistent if a plan that is optimal fromthepointofviewofoneperiodisnolongeroptimalfromthepointofview of a later period.” Our paper is concerned with so-called real options. This term, coined by Stewart Myers (Myers, 1977) indicates investments that offer options for future action and are thus characterized by several decision points in time, meaning a dynamic decision structure (cf. Dixit and Pindyck, 1994). Related literature refers to different options providing managerial flexibility (e.g. the option to abandon, contract, expand, stage, defer, grow or switch). Our focus is on the option to defer, or delay, an investment. The value of this option depends on the time of its “exercise”. Exercising suboptimally (where it is worth waiting or when the optimal exericse time has passed) reduces its value. This is where the problem of time inconsistency arises. Todate,onlyafewpapershavetackledtheissueoftimeinconsistencyinthe sense of quasi-hyperbolic discounting and options to defer an investment. The most important one, the study by Grenadier and Wang (2007), is concerned with risk-neutral decision makers that do not deal with (or have no access to) capital market interaction. A similar setting has been developed by Quah and Strulovici(2013)inthecontextofoptimalstopping. Clearly,theoptiontodefer an investment can be understood as a problem of optimal stopping, where the flowpayoffsarezeroandtheterminationpayoffequalsthebenefitsofexercising theoption. Yet,theassumptionofriskneutralityisquiteproblematicasexperi- mentalandempiricalevidenceshowsthatdecisionmakersarerisk-averse(foran overview, see Lengwiler, 2004, pp. 68–101). Moreover, following the economic principle,opportunitiestotradeonthecapitalmarketshouldnotbeoverlooked, as the devaluation of future results occurs not only due to impatience or pure time preference but also as a consequence of (market-based) opportunity costs (Mulligan, 1996; Read, 2004). Additionally, financial instruments permit the decisionmakertooptimallyspreadconsumptionovertimeandstatesaccording to their preferences (Smith and Nau, 1995; Smith, 1998). Therefore, the paper at hand is structured along two dimensions (see Ta- ble 1). The first one refers to the decision maker’s risk attitude and includes the assumptions of risk neutrality and risk aversion. The second one covers the extent to which capital market strategies are admissible. Here, distinction is madebetweenthecasesofnocapitalmarket,onlyrisk-freeborrowingandlend- ing, as well as spanning. In contrast to the continuous-time models developed by Grenadier and Wang (2007) and Quah and Strulovici (2013), the framework presented exhibits a time-discrete structure. In a way, it builds on the article of Khan et al. (2013). Their paper deals with related questions, yet mingles market-basedandpreference-basedvaluationinasomewhatdebatablemanner. We will return to Khan et al. (2013) in a later section. The remainder of the paper progresses as follows. The next section in- troduces our basic framework and reflects the impact of quasi-hyperbolic dis- 2 Table 1: Classification of subject-based literature admissible capital market strategies risk no capital incompletemarketin spanning attitude market the sense of (only) risk-free borrowing and lending risk Grenadier and Wang this paper this paper neutrality (2007), (scenario I) (scenario III) Quah and Strulovici (2013) risk this paper this paper this paper aversion (scenario IV) (scenario II) (scenario III) countingforcertaincombinationsofriskattitudeandadmissiblecapitalmarket strategies. InSectionthree,wepresentanumericalexampleandperformacom- parativestaticanalysis. Sectionfoursummarizesmajorfindingsandshowssome implications. 2 The model 2.1 Formalization of time (in)consistency Frequently,theconceptofstationarityisusedtoformalizetimeconsistency(e.g. FishburnandRubinstein,1982,p.681). Itisstatedasfollows. Provided(x,h1) denotes an outcome x to be received h1 periods ahead, the following relation holds. If (x,h1)(cid:31)(y,h2) then (x,h1+∆)(cid:31)(y,h2+∆) (analogous for ≺ as well as ∼). Thatmeans,acomparisonbetweentwotime-dependentoutcomesdependsonly on the difference h2−h1 between the points in time. If the points in time are advanced or deferred by the same amount ∆, the order of preference will be preserved. Experimental evidence suggests frequent violations of stationarity (e.g. Thaler, 1981; Benzion et al., 1989), i.e. time-inconsistent behavior. The most common way to model time-inconsistent preferences is to assume thatateachpointintime,thedecisionmakerappliesso-calledquasi-hyperbolic discounting in accordance with (∀t=0,1,...,T) T (cid:88) Ut(x ,x ,...,x )=U(x )+β δ(τ−t)U(x ) t t+1 T t τ τ=t+1 and 0<β,δ <1 (Phelps and Pollak, 1968; Laibson, 1997). Here, (x ,x ,...,x ) denotes a t t+1 T stream of cash flows. U(x ) represent identical functions of periodical cash τ flows. These utilities are “discounted” by the factors β and δ. In particular, t=0 implies T (cid:88) U0(x ,x ,...,x )=U(x )+β δτU(x ). 0 1 T 0 τ τ=1 3 It is easy to show that these preferences may lead to dynamic inconsistency. Imagine the choice between deterministic cash flows xˆ (τ > 0) and xˆ . As τ τ+1 of time t=0, the decision maker exhibits a preference for xˆ if and only if τ βδτU(xˆ )>βδτ+1U(xˆ ) (1) τ τ+1 or U(xˆ )>δU(xˆ ). (2) τ τ+1 Yet, from the point of view of t = τ, the decision maker will prefer xˆ if and τ only if U(xˆ )>βδU(xˆ ), (3) τ τ+1 which clearly contradicts the former. 2.2 Basic framework In the following, we set up a simple two-period model. The preferences of the decision maker are given by the intertemporal utility function 2 (cid:88) U0(x ,x ,x )=U(x )+β δτU(x ). 0 1 2 0 τ τ=1 Whereβ =1,thefamiliarconceptofexponentialdiscountingisobtained. β <1 results in quasi-hyperbolic discounting as described above. Concerning period- ical utility, we differentiate between the cases of risk neutrality U(x )=x τ τ 2 (cid:88) ⇒U0(x ,x ,x )=x +β δτx (RN) 0 1 2 0 τ τ=1 and (constant absolute) risk aversion U(x )=1−e−αxτ τ 2 ⇒U0(x ,x ,x )=1−e−αx0 +β(cid:88)δτ(cid:0)1−e−αxτ(cid:1).1 (RA) 0 1 2 τ=1 We want to analyze the following option to defer an investment. As shown in Figure 1, the benefits of exercising the option (modeled as a lump-sum payoff) follow a multiplicative binomial process, where b denotes the initial value. u and d represent growth factors that satisfy u > d and u = 1/d. p indicates the probability of an up-move. “Exercise”oftheoptionispossibleint=1ort=2atcostofinvestmentf. To simplify calculations, b=f is assumed. If the option is held until maturity at date t = 2, the decision maker will only exercise if two successive up-moves occur, leading to a cash flow of u2b−f. A premature exercise at t=1 may be advantageous after one up-move. The resulting cash flow of ub−f has to be weighed up against the consequences of deferral. The latter amount to u2b−f (two up-moves) and 0 (one up- and one down-move), respectively. 1According to Kirkwood (2004), an appropriately chosen exponential utility function closelyapproximatesgeneralutilityfunctionsinmostcases. 4 u2b p ub p 1 −p b b p 1 −p db 1 −p d2b t=0 t=1 t=2 Figure 1: Process of the benefits of exercising the option 2.3 Scenario I: Risk neutrality and risk-free borrowing and lending In the first scenario, the case of risk neutrality and risk-free borrowing and lending is analyzed. As a basis, we calculate the project value, meaning the multi-period certainty equivalent of the risky stream (given a particular state- contingent exercise strategy) to the decision maker. This is done in two steps. The first step deals with the situation of a single period, the second general- izes. The certainty equivalent is understood as a certain amount at the date of valuationwhoseutilityequalstheexpectedutilityoftheriskystreamwhenbor- rowing and lending opportunities are taken into account.2 Formally, for t = 0, we obtain max(ce−y)+βδ[(1+r)y] y =max(−z)+βδE[x +(1+r)z] 1 z where ce denotes the multi-period certainty equivalent and y as well as z rep- resent risk-free borrowing and lending over one period. r indicates the risk-free market rate. Inserting E[x ] ce= 1 1+r and y =z+ce (and ignoring issues of uniqueness if βδ (cid:54)= 1/(1+r)) yields max(ce−(z+ce))+βδ[(1+r)(z+ce)] z =max(−z)+βδ[E[x ]+(1+r)z] 1 z 2The literature on option pricing frequently uses the term utility indifference price (cf. HendersonandHobson,2009) 5 or max(−z)+βδ[E[x ]+(1+r)z] 1 z =max(−z)+βδ[E[x ]+(1+r)z] 1 z whichshowsthatvaluation(andthedeterminationoftheoptimalexercisestrat- egy) are preference-free and time inconsistency has no effect. Now the second step is quite trivial. Due to the properties of the expectation operator, i.e. lin- earity and the law of iterated expectations, this result can be generalized in an obvious way to multiple periods. An alternative derivation of our finding can be found in Hakansson (1969). As a consequence of the valuation result, the following exercise policy is obtained. The decision maker exercises prematurely afteroneup-moveifthenetbenefitsofimmediateexercisece exceedthevalue 1,e associated with waiting to invest ce , i.e. 1,n u2b−f ce =ub−f >ce =p . 1,e 1,n 1+r 2.4 Scenario II: Risk aversion and risk-free borrowing and lending The second scenario covers the combination of risk aversion and risk-free bor- rowing and lending. It is analyzed with recourse to the work of Smith (1998). To start with, note that the intertemporal utility function (RA) can be written in the form of 2 U0(x ,x ,x )=1−e−αx0 +β(cid:88)δτ(cid:0)1−e−αxτ(cid:1) 0 1 2 τ=1 2 2 (cid:88) (cid:88) = k − k e−αxτ, τ τ τ=0 τ=0 using utility weights k = 1, k = βδ and k = βδ2. Defining risk tolerances 0 1 2 ρτ =ρ¯=1/αandtakingcareoftheinvariancetopositiveaffinetransformations, we obtain 2 (cid:88) U0(x ,x ,x )=− k e−xτ/ρτ, (SM) 0 1 2 τ τ=0 which equals the preference funtion used by Smith (1998, p. 1969). Moreover, theabovebinomiallatticecanbetransferredintoadecisiontree(Figure2). As is customary, the boxes are called decision nodes and indicate decisions to be made. The circles are called chance nodes and represent the states relevant to the point of decision. Outcomes are written to the tips of the tree. Smith (1998) introduces the following rollback procedure to solve for the optimal policy (as of date t=0): (i) Calculate net present values (NPVs) for each endpoint in the tree by dis- counting all cash flows along the path leading to that endpoint using the risk-free rate 2 ce =(cid:88) xτ . 2 (1+r)τ τ=1 6 (payment in t = 1, payment in t = 2) 1 (ub−f,0) e x ercise noexercise p (0,u2b−f) 1 p −p (0,0) 1 1 − (db−f,0) p e x ercise noexercise p (0,0) 1 −p (0,0) t=0 t=1 t=2 Figure 2: Decision tree of the option exercise problem 7 (ii) At chance nodes, calculate so-called effective certainty equivalents using theexponentialutilityfunction(SM)witheffectiverisktoleranceR . For- τ mally, this involves (∀τ =1,2) (cid:16) (cid:104) (cid:105)(cid:17) ce =−R ln E e−ceτ/Rτ τ−1 τ 2 with R =ρ¯(cid:88)(1+r)−t, τ t=τ where ce denotes possible successor values. τ (iii) Atdecisionnodes,choosethemaximumvalueoftheavailablealternatives. (iv) The value at the root of the tree equals (in our terminology) the project’s multi-period certainty equivalent. As Smith (1998, p. 1697) points out, the multi-period certainty equivalent to the decision maker does not depend on the utility weights k . Consequently, τ there is no valuation effect of quasi-hyperbolic discounting. After one up-move, the rollback procedure stated translates into the following exercise policy. The decision maker exercises prematurely if ρ¯ (cid:16) (cid:17) ce =ub−f >ce =− ln pe−(u2b−f)/ρ¯+(1−p)e−0/ρ¯ . 1,e 1,n 1+r 2.5 Scenario III: Spanning Inthethirdscenario,weanalyzethecaseofspanning. Thismeansthatthecash flow of the project to be valued can be perfectly duplicated on the capital mar- ket (DeAngelo, 1981). Again, the project value is understood as multi-period certainty equivalent of the decision maker, i.e. a certain amount at the date of valuation, whose utility equals the expected utility of the risky stream (given a particular state-contingent exercise strategy) when trading opportunities in duplicating securities (including risk-free borrowing and lending)are taken into account. Formally, as of date t=0, this implies maxU(cid:0)ce−ψTP (cid:1)+βδE(cid:2)U(cid:0)(ψ −ψ )TP (cid:1)(cid:3)+βδ2E(cid:2)U(cid:0)ψTP (cid:1)(cid:3) 0 0 0 1 1 1 2 ψ0,ψ1 =maxU(cid:0)−ξTP (cid:1)+βδE(cid:2)U(cid:0)x +(ξ −ξ )TP (cid:1)(cid:3)+βδ2E(cid:2)U(cid:0)x +ξTP (cid:1)(cid:3), 0 0 1 0 1 1 2 1 2 ξ0,ξ1 whereψ ,ξ (∀τ =0,1)denotestate-contingenttradingstrategiesinduplicating τ τ securities(includingrisk-freeborrowingandlending). P (∀τ =0,1,2)represent τ vectors of state-contingent securities prices. Setting (∀τ =0,1) ξ =ψ −ψd, τ τ τ where ψd represents the state-contingent duplicating trading strategy, yields τ maxU(cid:0)ce−ψTP (cid:1)+βδE(cid:2)U(cid:0)(ψ −ψ )TP (cid:1)(cid:3)+βδ2E(cid:2)U(cid:0)ψTP (cid:1)(cid:3) 0 0 0 1 1 1 2 ψ0,ψ1 = maxU(cid:0)−[ψ −ψd]TP (cid:1)+βδE(cid:2)U(cid:0)x +([ψ −ψd]−[ψ −ψd])TP (cid:1)(cid:3) 0 0 0 1 0 0 1 1 1 ψ0,ψ1 +βδ2E(cid:2)U(cid:0)x +[ψ −ψd]TP (cid:1)(cid:3). 2 1 1 2 8 With the definition of the duplicating trading strategy, we obtain maxU(cid:0)ce−ψTP (cid:1)+βδE(cid:2)U(cid:0)(ψ −ψ )TP (cid:1)(cid:3)+βδ2E(cid:2)U(cid:0)ψTP (cid:1)(cid:3) 0 0 0 1 1 1 2 ψ0,ψ1 = maxU(cid:16)ψdTP −ψTP (cid:17)+βδE(cid:2)U(cid:0)(ψ −ψ )TP (cid:1)(cid:3)+βδ2E(cid:2)U(cid:0)ψTP (cid:1)(cid:3). 0 0 0 0 0 1 1 1 2 ψ0,ψ1 Obviously ce=ψdTP , 0 0 this means that the multi-period certainty equivalent of the decision maker equals the current value of the duplicating trading strategy irrespective of the shapeofperiodicalutility. Thevaluation(andthedeterminationoftheoptimal exercise strategy) are thus preference-free and there is also no effect of quasi- hyperbolic discounting. Aderivationoftheoptimalexercisepolicyrequiresthespecificationofavail- able capital market instruments. It is assumed that, in addition to the possi- bility of risk-free borrowing and lending at an interest rate r, a risky market instrument is traded. The price process of this security follows a multiplicative binomial process as shown in Figure 3. u¯ and d¯represent growth factors that satisfy u¯>d¯, u¯=1/d¯and u¯>1+r >d¯. u¯2 p u¯ p 1 −p 1 1 p 1 −p d¯ 1 −p d¯2 t=0 t=1 t=2 Figure 3: Price process of the risky market instrument With the help of this specification, the duplication of project cash flows can be condensed in the use of so-called risk-neutral probabilities 1+r−d¯ q = u¯−d¯ (for the up-move) and 1−q (for the down-move), respectively. Consequently, the decision maker exercises prematurely after one up-move if u2b−f ce =ub−f >ce =q . 1,e 1,n 1+r 9 An explicit instruction is obtained, if the “underlying” of the real option is traded or if the benefits of exercising the option exhibit the same dynamic as therisky(market)instrument,i.e.u=u¯. Inthiscase,asiswell-known(Merton, 1973, p. 144; Cox et al., 1979, pp. 234-235), a premature exercise of the option to defer is never optimal. 2.6 Scenario IV: Risk aversion and no capital market The fourth scenario covers the case of risk aversion without capital market interaction. In this context, the value of the project equals the multi-period certainty equivalent of the risky stream (given a particular state-contingent exercise strategy) at the date of valuation. From the point of view of t = 0, it is defined by U0(ce,0,0)=E(cid:2)U0(x ,x ,x )(cid:3) 0 1 2 or 2 1−e−αce =1−e−αx0 +β(cid:88)δτE(cid:2)1−e−αxτ(cid:3). τ=1 Consequently, after one up-move, we obtain the following exercise policy. A premature exercise results in a cash flow of ce =ub−f. 1,e In contrast, the expected utility of waiting to invest amounts to E(cid:2)U1(x ,x )(cid:3)=pβδ(cid:16)1−e−α(u2b−f)(cid:17). 1 2 With the condition ce =U−1(cid:0)E(cid:2)U1(x ,x )(cid:3)(cid:1), 1,n 1 2 this translates into 1 (cid:16) (cid:16) (cid:17)(cid:17) ce =− ln 1−pβδ 1−e−α(u2b−f) . 1,n α Clearly, the decision maker exercises after one up-move if ce >ce . 1,e 1,n Usingthisresult,itispossibletoelaborateontheeffectoftimeinconsistency. A time-inconsistent decision maker will exercise the option earlier than a time- consistent decision maker if 1 (cid:16) (cid:16) (cid:17)(cid:17) − ln 1−pβδ 1−e−α(u2b−f) α 1 (cid:16) (cid:16) (cid:17)(cid:17) <ub−f <− ln 1−pδ 1−e−α(u2b−f) α holds. Solving 1 (cid:16) (cid:16) (cid:17)(cid:17) − ln 1−pβδ 1−e−α(u2b−f) =! ub−f α for β yields 1−e−α(ub−f) β¯= . (cid:0) (cid:1) pδ 1−e−α(u2b−f) Ifβ ≤β¯(andβ¯<1)atime-inconsistentdecisionmakerwillexercisetheoption earlier (for a similar reasoning, Khan et al., 2013, p. 207). β¯ is thus called the criticalvalueofthepresent-biasself-controlparameter. Astheexpressionabove shows, β¯ is a function of α,u,b,f,δ and p. 10

Description:
(Phelps and Pollak, 1968; Laibson, 1997). Here, (xt,xt+1,,xT ) denotes . 2The literature on option pricing frequently uses the term utility indifference price (cf. Henderson and .. Theory of rational option pricing. Bell Journal of Eco-.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.